xCloth: Extracting Template-free Textured 3D Clothes from a Monocular Image

TL;DR

This paper introduces xCloth, a novel framework for template-free 3D garment reconstruction from monocular images that captures high-frequency textures and arbitrary clothing styles without human intervention.

Contribution

It extends PeeledHuman representation to predict layered depth and semantic maps, enabling UV parametrization and high-fidelity textured 3D garment digitization.

Findings

Achieves high-quality 3D garment reconstruction on multiple datasets.

Generalizes effectively to internet images.

Retains high-frequency textural details in reconstructed garments.

Abstract

Existing approaches for 3D garment reconstruction either assume a predefined template for the garment geometry (restricting them to fixed clothing styles) or yield vertex colored meshes (lacking high-frequency textural details). Our novel framework co-learns geometric and semantic information of garment surface from the input monocular image for template-free textured 3D garment digitization. More specifically, we propose to extend PeeledHuman representation to predict the pixel-aligned, layered depth and semantic maps to extract 3D garments. The layered representation is further exploited to UV parametrize the arbitrary surface of the extracted garment without any human intervention to form a UV atlas. The texture is then imparted on the UV atlas in a hybrid fashion by first projecting pixels from the input image to UV space for the visible region, followed by inpainting the occluded regions. Thus, we are able to digitize arbitrarily loose clothing styles while retaining high-frequency textural details from a monocular image. We achieve high-fidelity 3D garment reconstruction results on three publicly available datasets and generalization on internet images.